A version of an O2 Desktop Amp (ODA)

Last edited:
2000impreza - good work on the google link! I never could figure out how to "permalink" google drive so that it would stay up for more than a day.

The first R59, the 24.3k, should be removed completely. Also remove the 24V fixed voltage regulator (IC13) and C52. Those were all remnants of the change I made to power the relay circuit directly from the pre-regulator outputs.

Here is the latest revision of the build instructions. I've added more information and a couple of more sections.
 

Attachments

  • ODA BUILD INSTRUCTIONS 16.0.pdf
    252 KB · Views: 135
mounting an ODAC inside the amp

The new 16.0 build instructions include a section on mounting an ODAC inside the amp. There are two ways.

In the standard B4-080 box I've left space above the PC board. A 80x160mm perfboard or copper clad board can be inserted in the top slot, then an ODAC mounted upside down underneath. The new board will have to be pulled up just slightly to clear the 820uF caps. Since those have electrically live bare tops, be sure to put a piece of electrical tape over those first if using a copper clad board. Or if the board is clad on just one side flip it over so the non clad side is against the caps. A hole would need to be drilled or cut into the front panel for the USB port.

There is a chance the ODAC could fit above that copper clad board in the top slot of the case, between the board and the top of the case. I'm not sure since I can't find thickness measurements on the ODAC. If it fits just drill a hole in the clad board to run the audio wires down to the main PC board. This would be especially slick since that coppper clad layer would act like an EMI shield between the digital stuff on the ODAC and the sensitive analog parts on the main PC board.

The other way requires two of the PC boards and the longer B4-160 case. The two boards slide in end-to-end. Only the power supply parts are populated in the rear board while everything but the power supply parts are populated in the front board. Then run 3 twisted wires between the holes marked V+, V-, and GND on the two boards. Note to only one of the upper two holes in the GND hole section. A short USB cable would have to be used to cable the USB port out to a port on the back panel.

The second method with the 2 boards is kind of nice in that it puts all the noisy stuff in back, the AC part of the power supply and the digital parts of the ODAC, while leaving just the quiet analog signal stuff in front. The difference in cost between the B4-080BL box and the B4-160BL box is $11 at Allied electronics.

An external ODAC can also be used, of course, like JDS labs offering in its own case.
 
ODA beats O2 FFT in ARTA - and a design change: remove CLC coils

I have a design change to post in the amplifier. Don't install the two 47uH inductors in the power supply CLC filter. Instead just jumper both using a 20 or 22 gauge bare wire through the holes, or under the PC board with a straight wire.

I used unshielded coils and it turns out the power line harmonics radiate into the amplifier circuitry. Not enough to be audible, but they showed up on the FFT, below. If I make a future version someday I'll try to fit in some shielded coils.

The change just converts the filter from C-(LR)-C to CRC. The filter is there simply as extra protection from audio band noise in the AC wall socket from getting into the amplifier. The change has no effect at all on the performance of the amplifier.

Sorry about the screw up! I should have known better than to use unshielded inductors.



Now the good news. :D Using the ARTA soundcard software, the ODA's FFT beats the O2 below 1kHz, as shown below. I'm not a fan of soundcard analysis software as I've written many times. But it occured to me a few days ago that although soundcard software may be useless for absolute measurements, it may in fact have value for relative measurements. In other words, comparing one amp vs. another, ignoring the absolute numbers and instead just looking at relative numbers (the difference in the two measurements). Whatever distortions and noise the soundcard/soundchip adds will add to both and more-or-less subtract out in the difference. In this case it is comparing this amp vs. the O2, which of course has accurate published absolute numbers from RocketScientists's dScope measurements.

The first two photos show the test setup with the ODA and with the O2. That is a standard O2 that I've just mounted int a taller B3 case. The ARTA software

ARTA Home

in this case works by outputing the test signal on the PC's (24/96 DAC in PC) left channel out, which then goes to the amp's left (or right, just be consistant) input. Then the amp's left channel out goes to the PC's left channel in. I've standardized the output levels of the two amps for the test to -7.8dB. The ARTA software is running in 64 bit mode on windows 7 pro and the soundchip is set for 24/96.

The first spectrum photo below is with the two power supply 47uH inductors installed, and is what tipped me off to the radiated EMI problem. Those spikes you see left of the 1kHz fundamental are all power line harmonics, right on the money: 60Hz, 120Hz, 180Hz, 240Hz, etc. The cursor is sitting right on the 240Hz peak. The levels are all way too low to be audible, which is why I never noticed them when listening to the amp. So theoretically you could actually leave the inductors in, but I would highly recommend nuking them to get those radiated power line harmonics out of the system/spectrum.

The next spectrum photo is the ODA here with the inductors removed. Not bad, relatively! Now keep in mind the absolute values of dBFS shown on the Y axis are useless since the PC's sound chip's (significant, I've tried it in loopback) noise and distortion are being added, but what were are looking for here is relative change. In this case the power harmonics are gone without the coils.

The final (third) spectrum photo is the O2, with output adjusted for the same -7.8dB. ARTA shows much higher spectral components below the 1kHz tone, up to about -56dBFS vs. -74dBFS for the ODA. Interesting. I'm still pondering why. I was expecting the ODA's FFT to beat the O2 slightly, but this is more than expected. I'm going to try again to verify without the software, instead using a signal generator and the oscilloscope's FFT.

The final two pictures are photos that proto panel has sent of the panels being fabbed. I've picked two different fonts to give a choice. I'm making the panels self-grounding. Recall that RocketScientist/NwAvGuy required that a small wire go from the ground of the input jack around one of the panel cover screws to ground the case. I wanted to make that automatic, so I've countersunk around the input jack. The jack's nut (comes with the jack) then grounds the front panel to the input ground. A countersink around the bottom right screw grounds the panel to the case. A similar countersink in the back panel grounds that to the case.

The countersinks cut through the anodization so the nuts/screws make good electrical contact. Apply some aluminum anti-oxidant compound on the countersinks and it should remain a high quality ground. The other holes in the front panel I've slightly oversized to prevent any of the other shafts or jacks from touching, making the panel single point grounded at the 3.5mm input jack. With this arrangement no wires at all need to be run from the PC board to panel or case for grounding.
 

Attachments

  • IMG_1992.JPG
    IMG_1992.JPG
    215.2 KB · Views: 358
  • IMG_1994.JPG
    IMG_1994.JPG
    199.3 KB · Views: 342
  • IMG_1990.JPG
    IMG_1990.JPG
    108.4 KB · Views: 341
  • IMG_1987.JPG
    IMG_1987.JPG
    105.7 KB · Views: 336
  • IMG_1993.JPG
    IMG_1993.JPG
    123.7 KB · Views: 325
  • 20130802_115356.jpg
    20130802_115356.jpg
    792.6 KB · Views: 154
  • 20130730_170345.jpg
    20130730_170345.jpg
    500.6 KB · Views: 144
Last edited:
The blue panels looks really good!
I might steal that design from you if I ever get around to build one of these :D

Yeah I like how those turned out!

I'll be posting all the updated measurements for self-drilling along with the Proto Panel CAD files. With the CAD files it is easy enough to open them with the free "Cetina" program on Proto Panel's site and modify all sorts of stuff. Different colors for the lettering, italics and bold for the text, moving text around and different fonts. Proto Panel said the lettering looked good just with the raw engraving into the aluminum, without any color fill. I almost picked yellow for the lettering instead of white. Then it would have looked like Hakko's new line of soldering stations. :)

HakkoUSA - Products - Soldering
 
Final panel CAD files and dimensions

I received the latest version of panels today from Proto Panel and everything fits reasonably well. After seeing the panels in person I've decided that smaller font in the picture posted above in post #184 is too narrow, but the wider font in the other picture is great. So I'm only going to use the one with the wider font

The site here won't let me upload the .cgc CAD files - unknown file type - so until I get that figured out I'll just send the files to Proto Panel to have on file. Best to send them an email to order and reference the desktop headphone amp on DIY Audio. Below are the final set of dimensions for self-drilling, or loading into any other CAD system like Front Panel Express.

Proto Panel volunteered to offer a discount to DIYers on the forum on the panels. I'll get in touch with them next week and see what they have in mind. They have had 40% off specials the last couple of months, maybe they have that going on this month too. They are at:

Custom Front Panels, Instrument Panels Manufacturer, Proto Panel

Their "Cetina" program calculates costs with the rightmost menu button. With no discount it comes up with:

Front panel no lettering ...$17.52
Front panel with lettering..$31.37
Rear panel no lettering..$16.53
Rear panel with lettering...$27.16

And again you can simply self-drill the two panels that come with the B4-080 (or B4-160) case if you have a drill press and a good caliper for measuring. Or you can take the measurements below and load them into Front Panel Express' software or any other outfit that does panels. Proto Panel is just coming up as the best deal I've found at the moment and they seem to do good work.
 

Attachments

  • ODA panel CAD dimensions 5.pdf
    68.2 KB · Views: 67
  • ODA rear panel.PNG
    ODA rear panel.PNG
    55.4 KB · Views: 90
  • ODA front panel.PNG
    ODA front panel.PNG
    65.4 KB · Views: 117
Last edited:
Panel CAD files posted now & better O2 FFT

I just discovered that the Google Drive link I posted back in #16 is working now for some reason. The mysteries of Google. So problem solved on posting the panel CAD files for Proto Panel. Go to that link

https://drive.google.com/folderview?id=0B67cJELZW-i8VmhVNk5PODNtZnc&usp=sharing

Then to 80x160mm -> 8_11_2013 -> Front and Rear Panels. Then just download the .cgc files for the front and back panels, labeled or unlabelled. To do that click on the file and Google will complain about "no preview available" since it doesn't know .cgc files. But there will be a download link in the lower right corner. Click on that and choose "save file".

Then just go to the ordering page on Proto Panels website:

Proto Panel Order Section

and when it asks to upload files, upload the front panel .cgc file for the first one and the rear panel .cgc for the second. No need to email them.

I've already made a couple of small changes in the measurements after bolting down the panels and spending some time with it. The latest set of measurements is below and in that Google link. I've decided it looks a little cooler with the black nylon nut screwed on the 1/4" Neutrik jack. That nut is an option available at Mouser and I'll add it to the BOM. But the nut covers up the 1/4" lettering so I've just removed that entirely. Now that jack just says "output". Then I've enlarged the holes around the RCA jacks by 0.5mm after trying a number of RCA cables. A few flare out at the ends slightly and the change prevents those from touching the case. Then finally I've enlarged the countersink around the panel corner screws 0.5mm to account for hole variances.

Pictures of the assembled panels are below. A few assembly notes..

Use the finer pitch 4mm Torx "thread forming" screws packed with the B4-080 case rather than the coarser pitch "thread cutting" screws they also pack. The thread forming can be taken out a few times without damage and will leave fewer metal filings. However, oddly enough, Box Enclosures only includes 8 of each screw with the case but it has 5 holes in each panel, including the one at the bottom. That leaves it 2 screws short. I have an email into them about it. But they have the screws available on their website for order, the TB-008

Box Enclosures - New Page

A #10 Torx driver is needed to get those screws in.

The black nylon nut for the Neutrik jack I mentioned above is Mouser #550-1005 (Neutrik #NRJ-NUT-B) for $0.14. It isn't required - the jack is soldered on the PCB and is perfectly stable - I just think it looks kind of slick. :p

I've found these rubber feet work well, Philmore 10-602 available from all sorts of places on the internet. Here they are at Fry's

FRYS.com*|*Philmore

The adhesive sticks and stays, unlike a few less-rubbery feet I've tried, plus it has a hole in the center and could be mounted with a #8 self tapping screw.

In the photos I've replaced the Alpha DP4T rotary gain switch with the Alps version, from the discussion a few posts back in this thread. The Alps switch works just fine and allows that same T-8 knob used on the volume switch to be used, as I've done in these photos. I have the Alps switch as the primary part on the BOM now, with the Alpha version out in the notes as an option.

The mounting screw that goes into the RCA jacks is a #4 3/8" pan head. It doesn't come with the jack, but is pretty readily available at hardware stores. A good place I've found for hardware on the net is Bolt Depot:

Bolt Depot - Nuts and Bolts, Screws and Fasteners online

The M3 metric stainless button head Allen screws, nuts, and washers I'm using to hold the 4 voltage regulators are from there. I'm going to add info on the BOM for all these mounting hardware parts.

I forgot to post a link to the aluminum anti-oxidant compound I had in mind to use on the countersinks. This stuff or the equivalent, Ideal Industries NoAlOx:

Noalox Anti-Oxidant Compound

Home Depot here in the US has it, plus a bunch of places online. Electricians use it all the time on aluminum wiring, so any place that has residential or commercial electrical switches and panels should have it.

And the final couple of screen captures are new updated FFTs of the ODA (first one) and O2 second one. My FFT in the post above must have had a cable issue on my O2 test. I haven't been able to reproduce that higher noise plot on the O2 prior to the 1kHz. I also forgot to put a load on the O2 in that previous FFT - the ODA has a built-in minimum load - so both of these are with a 32R load. I've tried several tests since and this is what I get, which looks just neck and neck between the O2 and this version of an ODA amplifier. Again the absolute numbers are useless due to PC soundchip noise, just looking at the difference between the two, which looks negligible here.

I've applied averaging this time to both FFT's. Note that both the ODA and O2 still show tiny amounts of powerline harmonics below 1Khz. They are below audible levels and are a result of the half wave power supply parts being nearby to the amplifier parts in both amps. As time permits I'm going to build up and ODA in the longer B4-160 case with "two PCB" option that has the power supply parts in the back and see if that reduces the powerline stuff. I also did try a signal generator instead of the ARTA 1kHz feed, at the same 138mV level, but the delta function wasn't any better (less wide 1Khz fundamental).
 

Attachments

  • ODA panel CAD dimensions 6.pdf
    68.2 KB · Views: 47
  • IMG_2003.JPG
    IMG_2003.JPG
    172.9 KB · Views: 118
  • IMG_2000.JPG
    IMG_2000.JPG
    189.2 KB · Views: 120
  • IMG_2001.JPG
    IMG_2001.JPG
    194.3 KB · Views: 127
  • ODA FFT AVERAGE.jpg
    ODA FFT AVERAGE.jpg
    115.3 KB · Views: 120
  • O2 FFT AVERAGE.jpg
    O2 FFT AVERAGE.jpg
    116.2 KB · Views: 83
Last edited:
2 tle2064 for quad parallel output and still low power/portable as O2 ?

The TLE2064, or the economy version TL064 that I use in this amp for the clipping indicator (but isn't in the actual audio path), are unfortunately older chips that would be pretty bad for audio. Take a look at the THD graph on page 55 in the data sheet - pretty scary:

http://www.ti.com/lit/ds/symlink/tle2064bm.pdf

That chip would be most useful for non-audio things, like instrumentation amps, strain gauges, or level detectors in the case of this amp. You would be better off with a more modern quad chip that is audio-specific with better AC parameters (THD, noise, etc.) like the LME49740 or OPA1664. Both are cheaper, too, about $3.50 at Mouser vs. $4.50 or so for that TLE2064.

http://www.ti.com/lit/ds/symlink/lme49740.pdf

http://www.ti.com/lit/ds/symlink/opa1664.pdf

Even then you could only power higher impedance phones, like 150 ohm and above, due to the low current output. And with a quad chip you would have to keep an eye on the math and make sure the total power dissipation of the package doesn't get exceeded. Going to 2 dual chips would give twice the power dissipation.

For an extreme version of that, see this thing I came up with: :p

http://www.diyaudio.com/forums/headphone-systems/237041-massively-parallel-lme49990-headamp.html

But vs. the O2, RocketScientist/NwAvGuy seems to have come up with an optimum solution for a portable amp using that one dual NJM4556A. It is a high current output chip that still has reasonable AC parameters, at least when used as a unity gain buffer. Keep in mind that if you use a quad chip you also get hit with 4 times the quiescent current (idle current) in each chip section, and that in turn subtracts from the total remaining package power dissipation.
 
Last edited:
Looks like some high order powerline harmonics in the O2 amp

I just spotted something interesting in the O2 and ODA FFTs I posted above in #189. I can quite audibly hear that this ODA version's background is dead silent at the highest gain and volume vs. just the tiny edge-of-hearing hiss in the O2 with sensitive cans. I've been staring at the FFTs for the last hour or so puzzling over why, given how similar the two FFTs seem to be.

Two interesting things popped up. Below the 1kHz fundamental, from about 300Hz to 1kHz, the ODA has a lower average amplitude, by a couple of dB. But I don't think that is what I'm hearing in the O2 since the hiss sounds higher frequency.

But stare really hard at the O2 spectrum between 1kHz and about 4kHz. Although the average baseline level of the O2 is a couple of dB less that this ODA version in the entire portion to the right of 1kHz, see those tiny spikes from 1KHz to 4kHz? Measuring over and plotting the frequency those are high order powerline harmonics. 10th, 11th, and 12th harmonics - 1200Hz, 1800Hz, 2400Hz, etc. - that are all absent on the ODA spectrum to the right of the 1kHz delta. Looks like I managed to nuke those. Many headphones and ears

http://en.wikipedia.org/wiki/Fletcher%E2%80%93Munson_curves

would have much higher sensitivity to powerline harmonics in that 1-4kHz range than those below 1kHz.

Probably due to the close proximity of the power supply section in the O2 amplifier to the output amp parts. In this ODA version I had the luxury of a bigger case and more PC board real estate that RocketScientist/NwAvGuy didn't with the portable O2. I was able to put more spacing between the noisy AC parts and the amp parts.

So, in other words, that very slight edge-of-hearing background hiss in the O2 while on the highest gain and volume settings with input grounded and running on AC might actually be due to high order powerline harmonics, not atmospheric noise or chip/resistor noise.
 
Last edited:
ethanolson - you are right! :) I was thinking 120Hz for some reason. 1200Hz would be the 20th and 1800 would be the 30th harmonic. Thanks for catching that.

The Fluke 43B power quality meter can actually read down to the 51st harmonic:

Fluke 43B Single Phase Power Quality Analyzer

For electrical (buildings) use usually most of the power goes into the first 10 when something goes wrong. But for electronic purposes, even though the levels are way down there, the higher harmonics can figure into small amounts of noise.

Interesting story with the electrical stuff. A few years ago I was involved in tracking down some spooky random behavior with an electronic lighting controller system in a commercial building in los angeles. I noticed that many of the dry transformers were running hot, much more than they should given the rated loads, which is often a sign of harmonics. So I took power quality readings and sure enough, high levels of harmonics up to about the 8th. After a bunch of circuit tracing it turned out the entire building's ground had never been connected at the LADWP service entrance from the street. Oops. :) Once that was fixed all the building wiring harmonics went back down to normal levels and the lighting stuff worked properly again.
 
Last edited:
40% off at Proto Panel again - and updated BOM

It looks like Proto Panel is running their 40% off deal again through August. Here is the link with the discount code:

Custom Front Panels Milling 50% Discount

So that would cut the prices I posted above that their CAD program comes up with by 40% for either the panels with or without text.

Again, I 'm not affiliated with these folks in any way, it just seems like a good deal on panels and they have done great work on what I've ordered. The panels that come with the B4-080 case can simply be self-drilled with the measurements I've posted. I've kept everything as a round hole. At one point I was playing around with the idea of moving the power jack foward to be flush with the outside of the rear panel, but that would have required a square hole that is less DIY-friendly.

I have also posted an updated BOM below and at the Google link. I've added the part numbers for the knobs, switch buttons, rubber feet, RCA jack mounting screw, Neutrik 1/4" nylon nut, extra case bolts (both the black and silver) at Box Enclosures and voltage regulator (TO-220) mounting hardware. I've also included part numbers at Bolt Depot (boltdepot.com) in case anyone wants to use the metric M3 stainless steel button head parts I used instead of the 4-40 hardware that comes with most TO-220 mounting kits. The inductors are removed from the BOM since those are left off the board now. The relay is moved to the first section with the switches. I've included part numbers for different color versions of the knobs and buttons in the notes column of the BOM.

I found a mistake on the BOM. The relay should be a 48Vdc unit. It still had the 24Vdc unit listed, before I made the change to run the relay off the pre-regulators.

I've posted both an excel and PDF version at the Google link.
 

Attachments

  • ODA Headamp 10 PCB 80x160 BOM 72.pdf
    151.7 KB · Views: 122
Last edited:
Hopefully one or more of the O2 amp builders out there will eventually stock some parts!

I'm probably going to be offering a bag of the TO-220 mounting hardware in the vendor's forum using the stainless steel button-head socket (Allen wrench) metric 3mm stuff I used, plus those two mounting screws for the RCA jacks in stainless. Those are the optional part numbers from BoltDepot.com I have listed in the BOM. I'll be posting something in the vendor forum thread in a week or so when a bolt order arrives. I think the stainless stuff looks a little nicer than the standard zinc 4-40 bolts that come with the TO-220 mounting kit from Mouser that is in the BOM.

Someone reminded me that I can simply zip those front & rear panel CAD files and get them posted that way, so here they are. Same thing that is out at the Google link. Keep in mind you can open these up with Proto Panel's Cetina program and modify anything you want - font, colors, wording, text placement, etc.
 
Last edited by a moderator:
Hi Alex,

I just added a set of instructions for sending a PC board run off to Seeed Studio to that Google Drive link, in the "Gerber File" folder:

https://drive.google.com/folderview?id=0B67cJELZW-i8VmhVNk5PODNtZnc&usp=sharing

Essentially you download the Gerber .zip file from the Google link (or from post #173 here), rename it to something shorter, then go out to the Seeed 4 Layer service page here

Fusion PCB Service - 4 Layers [PCB02281P] - $39.90 : Seeed Studio Bazaar, Boost ideas, extend the reach

and enter that same .zip file in the upload box. Then just fill in the rest of the options - details in that Google Drive file.

For anyone doing a board run, do consider that 2-board build option in the longer B4-160BL case. The longer case is only about $5 more at newark.com, $26.50 vs. $21 for the B4-080BL, and the shipping would be about the same. The two boards just go end-to-end in the next-to-the-bottom slot. Then only populate the power supply stuff on the rear board and everything but the power supply parts on the front board. That leaves plenty of space on the rear board for mounting RocketScientist/NwAvGuy's ODAC. The same front and rear panels fit either case. A hole would need to be drilled in the rear panel of the longer case to get the USB cable out, and/or mount a USB connector for the ODAC.
 
Last edited: